Artificial intelligence-based automatic seed potatoes cutting device and method

Abstract

The present invention relates to an artificial intelligence-based automatic seed potato cutting device and method, wherein the device comprises: an image generation unit that photographs a seed potato to be cut and generates a seed potato image; a seed potato eye recognition unit that recognizes eyes in the seed potato image through a seed potato eye recognition model; a cutting angle calculation unit that calculates a cutting angle capable of cutting the seed potato while avoiding the eyes of the seed potato; and a cutting operation unit that performs a cutting operation of the seed potato according to the cutting angle. Accordingly, the present invention can increase the mechanization rate of potato planting operations, thereby replacing labor in agriculture and securing a stable potato production volume.

Description

Technology Field

[0001] The present invention relates to a seed potato cutting automation technology, and more specifically, to an artificial intelligence-based automatic seed potato cutting device and method capable of automatically cutting seed potatoes by considering the location and number of seeds recognized through a seed seed recognition model. Background Technology

[0003] Currently, the consumption of potatoes in Korea is increasing due to the Westernization of dietary habits. However, as the number of farmers decreases due to the aging population and the proportion of farmers aged 60 or older increases, the quantitative and qualitative decline in agricultural labor has led to a decrease in the potato cultivation area and self-sufficiency rate, resulting in reliance on imports for consumption.

[0004] In particular, during the planting stage of potato cultivation, seed potatoes are cut according to size before planting to increase yield. When cutting the seed potatoes, the cuts must be made to avoid the eyes while ensuring that at least one eye (sprout) exists on each piece. Additionally, all seed potatoes must be cut within a limited time to meet the planting schedule.

[0005] However, due to these demanding conditions, the process of cutting potatoes before planting is still carried out manually. This is attributed to the complexity of the seed potato planting operation, which requires the eyes to remain within the cut slices without damaging them. As of 2022, the detailed mechanization rates by potato work stage are 99.90% for tillage and soil preparation, 83% for plastic mulching, 6.20% for planting, 97.20% for pest control, and 71.90% for harvesting, indicating a significantly low mechanization rate in the planting sector.

[0006] Accordingly, mechanization of the planting stage is required for stable potato production, and machines that automatically cut seed potatoes are being developed. However, since all of them cut without considering the eyes of the seed potatoes, there is a disadvantage that the potato spoilage rate is high and the final sprout emergence rate is low.

[0007] Recently, there have been efforts to increase agricultural productivity through research on the application of artificial intelligence technology and unmanned automation. Accordingly, there has been a demand for the development of technology capable of mechanizing potato planting operations by applying AI. Prior art literature

[0009] Korean Registered Patent No. 10-1937226 (2019.01.04) The problem to be solved

[0010] One embodiment of the present invention aims to provide an artificial intelligence-based automatic seed potato cutting device and method capable of automatically cutting seed potatoes by considering the location and number of seeds recognized through a seed seed recognition model.

[0011] One embodiment of the present invention aims to provide an artificial intelligence-based automatic seed potato cutting device and method that learns seed germ image data to generate a seed germ recognition model, generates an optimal cutting line that cuts the seed potato while avoiding the recognized seed germ through the seed germ recognition model, and transmits the optimal cutting line to a seed potato cutter so that the cutting operation can be performed automatically.

[0012] One embodiment of the present invention aims to provide an artificial intelligence-based automatic seed potato cutting device and method that can increase the final sprout emergence rate of potatoes and reduce costs by automatically cutting the potatoes while considering the eyes before planting, and can contribute to replacing labor in agriculture and securing stable potato production by increasing the mechanization rate of planting operations. means of solving the problem

[0014] Among the embodiments, the artificial intelligence-based automatic seed potato cutting device comprises: an image generation unit that photographs a seed potato to be cut and generates a seed potato image; a seed potato eye recognition unit that recognizes eyes in the seed potato image through a seed potato eye recognition model; a cutting angle calculation unit that calculates a cutting angle capable of cutting the seed potato while avoiding the eyes of the seed potato; and a cutting operation unit that performs a cutting operation of the seed potato according to the cutting angle.

[0015] The image generation unit above can perform at least one preprocessing operation regarding the seed potato image.

[0016] The above seed potato eye recognition unit can distinguish and display the eyes of the seed potato as bounding boxes on the seed potato image using a Mask R-CNN-based recognition model (Segmentation Model).

[0017] The above seed potato eye recognition unit can pre-construct the seed potato eye recognition model by using an open-source based mask sensor to label the eyes as polygons on the seed potato image and training the labeled custom dataset on a mask R-CNN.

[0018] The above cutting angle calculation unit can calculate the cutting angle using a cutting line selected based on the number of cases where an eye exists in the cutting piece among the cutting lines passing between eyes recognized based on the center of the seed potato image.

[0019] The above cutting angle calculation unit can generate a first straight line connecting the center of the seed potato image and the center of the recognized eye, generate a second straight line as an intermediate angle between the first straight lines, select up to two of the second straight lines that do not pass through the recognized eye, determine at least one of the two selected second straight lines as the optimal cutting straight line according to preset conditions, calculate the angle of the determined optimal cutting straight line as the cutting angle, and transmit it to the cutting operating unit.

[0020] The above cutting angle calculation unit determines both of the selected maximum two second straight lines as the optimal absolute straight lines when there is at least one eye in each of the four pieces cut using the selected maximum two second straight lines, and when there is no eye in at least one of the four pieces cut, it selects one of the selected maximum two second straight lines, and when there is at least one eye in each of the two pieces cut using the selected one second straight line, it determines the selected one second straight line as the optimal absolute straight line, and when there is no eye in at least one of the two pieces cut, it can indicate that cutting is impossible.

[0021] The above cutting operating unit controls the operation of the blade at the cutting angle to perform the cutting operation of the seed potato and can repeat the cutting operation according to the number of cutting angles.

[0022] The above cutting operating unit may include a conveying module for conveying the seed potato; and a cutting module comprising a replaceable blade for cutting the seed potato, a first driving means for rotating the blade according to the absolute angle, and a second driving means for vertically driving the blade.

[0023] Among the embodiments, the artificial intelligence-based automatic seed potato cutting method comprises: an image generation step of generating a seed potato image by photographing a seed potato to be cut through a camera; a seed potato eye recognition step of recognizing eyes in the seed potato image through a seed potato eye recognition model; an absolute angle calculation step of calculating a cutting angle that can cut the seed potato while avoiding the eyes of the seed potato; a conveying step of conveying the seed potato under a blade; and a cutting operation step of rotating the blade according to the cutting angle to adjust the angle and introducing the blade perpendicularly into the seed potato to perform a cutting operation.

[0024] The above seed potato eye recognition step can distinguish and display the eyes of the seed potato as bounding boxes on the seed potato image using a Mask R-CNN-based recognition model (Segmentation Model).

[0025] The above cutting angle calculation step can calculate the cutting angle using a cutting line selected based on the number of cases where an eye exists in the cutting piece among the cutting lines passing between the eyes recognized based on the center of the seed potato image. Effects of the invention

[0027] The disclosed technology may have the following effects. However, this does not mean that a specific embodiment must include all of the following effects or only the following effects; therefore, the scope of the rights of the disclosed technology should not be understood as being limited by this.

[0028] An artificial intelligence-based automatic seed potato cutting device and method according to one embodiment of the present invention can automatically cut seed potatoes by considering the location and number of seeds recognized through a seed seed recognition model.

[0029] An artificial intelligence-based automatic seed potato cutting device and method according to one embodiment of the present invention learns seed eye image data to generate a seed eye recognition model, generates an optimal cutting line that cuts the seed potato while avoiding the recognized seed eyes through the seed eye recognition model, and transmits the optimal cutting line to a seed potato cutter so that the cutting operation can be performed automatically.

[0030] An artificial intelligence-based automatic seed potato cutting device and method according to one embodiment of the present invention can increase the final sprout emergence rate of potatoes and reduce costs by automatically cutting the seed potatoes while considering the eyes before planting, and can contribute to replacing labor in agriculture and securing a stable potato production volume by increasing the mechanization rate of planting operations. Brief explanation of the drawing

[0032] FIG. 1 is a drawing illustrating an artificial intelligence-based automatic seed potato cutting system according to one embodiment of the present invention. Figure 2 is a diagram illustrating the system configuration of the automatic seed potato cutting device of Figure 1. Figure 3 is a diagram explaining the functional configuration of the automatic seed potato cutting device of Figure 1. FIG. 4 is a flowchart illustrating an example of an artificial intelligence-based automatic seed potato cutting process according to the present invention. Figures 5a-5b illustrate a seed potato eye recognition model using a mask R-CNN according to the present invention. FIG. 6 is a flowchart illustrating an optimal cutting angle calculation algorithm according to the present invention. Figure 7 is a diagram showing the cutting line calculated through the optimal cutting angle calculation algorithm of Figure 6. FIG. 8 is a drawing illustrating the mechanical configuration of an automatic seed potato cutting device according to one embodiment of the present invention. Figure 9 is a drawing showing the blade portion in Figure 8. Figure 10 is a diagram showing the operation process of the automatic seed potato cutting device shown in Figure 8. Specific details for implementing the invention

[0033] The description of the present invention is merely an example for structural or functional explanation, and therefore the scope of the present invention should not be interpreted as being limited by the examples described in the text. That is, since the examples are subject to various modifications and may take various forms, the scope of the present invention should be understood to include equivalents capable of realizing the technical concept. Furthermore, the objectives or effects presented in the present invention do not imply that a specific example must include all of them or only such effects; therefore, the scope of the present invention should not be understood as being limited by them.

[0034] Meanwhile, the meaning of the terms described in this application should be understood as follows.

[0035] Terms such as "first," "second," etc., are intended to distinguish one component from another, and the scope of rights shall not be limited by these terms. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0036] When it is stated that one component is "connected" to another component, it should be understood that it may be directly connected to that other component, or that there may be other components in between. Conversely, when it is stated that one component is "directly connected" to another component, it should be understood that there are no other components in between. Meanwhile, other expressions describing the relationships between components, such as "between" and "exactly between," or "adjacent to" and "directly adjacent to," should be interpreted in the same way.

[0037] A singular expression should be understood to include a plural expression unless the context clearly indicates otherwise, and terms such as "include" or "have" are intended to specify the existence of the implemented features, numbers, steps, actions, components, parts, or combinations thereof, and should be understood not to preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0038] In each step, identifiers (e.g., a, b, c, etc.) are used for convenience of explanation and do not describe the order of the steps; the steps may occur differently from the specified order unless a specific order is clearly indicated in the context. That is, the steps may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

[0039] The present invention may be implemented as computer-readable code on a computer-readable recording medium, and the computer-readable recording medium includes all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. Additionally, the computer-readable recording medium may be distributed across networked computer systems, so that computer-readable code can be stored and executed in a distributed manner.

[0040] Unless otherwise defined, all terms used herein have the same meaning as generally understood by those skilled in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having meanings consistent with the context of the relevant technology and should not be interpreted as having an ideal or overly formal meaning unless explicitly defined in this application.

[0042] FIG. 1 is a drawing illustrating an artificial intelligence-based automatic seed potato cutting system according to one embodiment of the present invention.

[0043] Referring to FIG. 1, the seed potato automatic cutting system (100) may include a camera sensor (110), a seed potato automatic cutting device (130), and a database (150).

[0044] The camera sensor (110) can photograph a seed potato to generate a seed potato image. Here, the camera sensor (110) may be implemented as a webcam, but is not necessarily limited thereto. The camera sensor (110) can photograph a seed potato to be cut and provide the seed potato image to the automatic seed potato cutting device (130). The camera sensor (110) may be connected to the automatic seed potato cutting device (130) via a network.

[0045] In one embodiment, the camera sensor (110) may correspond to a component of an automatic cutting system that performs automatic seed potato cutting operations and may be directly connected to an automatic seed potato cutting device (130).

[0046] The automatic seed potato cutting device (130) can perform the seed potato cutting process during the seed potato planting operation. The automatic seed potato cutting device (130) may be implemented by including a computer or program that performs automated seed potato cutting using a seed potato cutter. In one embodiment, the automatic seed potato cutting device (130) can recognize the eyes of the seed potato through a camera sensor (110) to calculate a cutting angle and perform the cutting operation of the seed potato using the calculated cutting angle. Here, the automatic seed potato cutting device (130) may be implemented by including a seed potato cutting machine such as a blade or a motor.

[0047] The automatic seed potato cutting device (130) can acquire an image of a seed potato to be cut from a camera sensor (110), input it into a pre-established seed eye recognition model to recognize the seed eye, and calculate a cutting angle to cut the seed potato while avoiding the seed eye. The automatic seed potato cutting device (130) can perform the seed potato cutting operation by controlling the operation of the seed potato cutter with the calculated cutting angle.

[0048] The database (150) may correspond to a storage device that stores various information required during the operation of the automatic seed potato cutting device (130). For example, the database (150) may store seed detection images captured by the camera sensor (110), may store learning algorithms and datasets for building an AI model for seed eye recognition, and may store algorithms for calculating the optimal cutting angle based on the position and number of recognized eyes. It is not necessarily limited to these, but may store information collected or processed in various forms during the process in which the automatic seed potato cutting device (130) performs artificial intelligence-based automatic seed potato cutting according to the present invention.

[0050] Figure 2 is a diagram illustrating the system configuration of the automatic seed potato cutting device of Figure 1.

[0051] Referring to FIG. 2, the seed potato automatic cutting device (130) may include a processor (210), memory (230), user input / output unit (250), and network input / output unit (270).

[0052] The processor (210) can execute a procedure to process each step in the process of operation of the automatic seed potato cutting device (130), manage memory (230) that is read or written throughout the process, and schedule the synchronization time between volatile memory and non-volatile memory in memory (230). The processor (210) can control the overall operation of the automatic seed potato cutting device (130) and is electrically connected to memory (230), user input / output unit (250), and network input / output unit (270) to control the data flow between them. The processor (210) can be implemented as the CPU (Central Processing Unit) of the automatic seed potato cutting device (130).

[0053] The memory (230) may include an auxiliary storage device that is implemented as non-volatile memory such as an SSD (Solid State Drive) or HDD (Hard Disk Drive) and is used to store all the data required for the seed potato automatic cutting device (130), and may include a main memory device that is implemented as volatile memory such as RAM (Random Access Memory).

[0054] The user input / output unit (250) may include an environment for receiving user input and an environment for outputting specific information to the user. For example, the user input / output unit (250) may include an input device including an adapter such as a touch pad, a touch screen, a virtual keyboard, or a pointing device, and an output device including an adapter such as a monitor or a touch screen. In one embodiment, the user input / output unit (250) may correspond to a computing device connected via remote access, and in such case, the seed potato automatic cutting device (130) may be performed as an independent server.

[0055] The network input / output unit (270) includes an environment for connecting to an external device or system through a network, and may include an adapter for communication such as a LAN (Local Area Network), MAN (Metropolitan Area Network), WAN (Wide Area Network), and VAN (Value Added Network).

[0057] Figure 3 is a diagram explaining the functional configuration of the automatic seed potato cutting device of Figure 1.

[0058] Referring to FIG. 3, the automatic seed potato cutting device (130) may include an image generation unit (310), a seed potato eye recognition unit (330), a cutting angle calculation unit (350), a cutting operation unit (370), and a control unit (390).

[0059] The image generation unit (310) can generate a seed potato image by photographing the seed potato to be cut. Here, the image generation unit (310) can generate a seed potato image by photographing the seed potato to be cut through a camera sensor (110). The camera sensor (110) can photograph the seed potato during the process of the seed potato being transported to the automatic cutting device (130), or it can be installed on a seed potato placement platform to photograph the seed potato placed on the platform immediately before transport. In one embodiment, a webcam may be used as the camera sensor (110), but it is not necessarily limited thereto. For example, a multi-RGB camera or a multi-RGB-D camera may be used as the camera sensor (110) to obtain an image with higher accuracy regarding the seed potato to be cut.

[0060] In one embodiment, the image generation unit (310) can receive an image of a seed potato to be cut, captured by a camera sensor (110), and can store and manage it in a database (150). The image generation unit (310) can perform at least one preprocessing operation regarding the seed potato image. For example, the image generation unit (310) can selectively perform preprocessing operations such as image filtering, grayscale, binarization, enlargement and reduction, rotation and transformation.

[0061] The seed potato eye recognition unit (330) can recognize the eyes of the seed potato by inputting the seed potato image into a pre-established seed eye recognition model. Here, the seed eye recognition model may correspond to a Mask R-CNN-based identification model (Segmentation Model) implemented to recognize seed eyes of varying shapes, sizes, and positions on the seed potato image by learning a seed potato image-based dataset. Since the shape, size, and position of the eyes differ from one seed potato to another, the eyes must first be recognized finely and accurately at the pixel level before cutting, taking the eyes into account. The seed potato eye recognition unit (330) may use Mask R-CNN to recognize the eyes.

[0062] Mask R-CNN is a type of object segmentation algorithm that is a deep learning-based algorithm capable of not only performing object recognition but also obtaining object masks. Compared to semantic segmentation, which performs a similar role, it has the advantage of being able to classify objects of the same type as distinct objects even if they overlap. Mask R-CNN is a deep learning algorithm training model that performs two tasks simultaneously. The first is 'object detection,' which detects objects based on pre-set object recognition criteria and generates their bounding boxes; the second is 'object segmentation,' which applies pixel-level masks to objects and divides them individually. Unlike YOLO, a model that recognizes objects using rectangular bounding boxes, Mask R-CNN recognizes objects by applying pixel-level masks, allowing it to learn by precisely extracting data based on the shape of the eye.

[0063] In one embodiment, the seed potato eye recognition unit (330) can distinguish and display the eyes of the seed potato on the seed potato image using a mask R-CNN-based recognition model. Here, the recognition model may correspond to an in-depth learning model implemented to classify which class each pixel of the image belongs to. That is, the seed potato eye recognition unit (330) can recognize the eyes on the seed potato image by constructing a recognition model in advance and utilizing it. Specifically, the recognition model can receive the seed potato image as input and generate an output image in which the recognized eyes on the seed potato image are distinguished and displayed using bounding boxes. If the seed potato image contains multiple eyes, the recognition model can generate an output image by displaying an independent bounding box for each recognized eye.

[0064] In one embodiment, the seed potato eye recognition unit (330) may operate in conjunction with a separate learning server, and in this case, the seed potato eye recognition for the seed potato image can be performed using a seed potato eye recognition model built by the learning server. Additionally, the seed potato eye recognition model may be built as a result of learning an image-based dataset for each seed potato. The seed potato eye recognition unit (330) can recognize the seed potato eye from the seed potato image through the previously built seed potato eye recognition model. If the seed potato eye recognition unit (330) cannot recognize the seed potato eye, it may output an eye recognition failure notification message.

[0065] The cutting angle calculation unit (350) can calculate the cutting angle based on the eyes of the recognized seed potato and transmit the calculated cutting angle information to the cutting operation unit (370) via serial communication. Here, the cutting angle information refers to the angle of the cutting line that avoids the recognized eyes. The cutting angle calculation unit (350) can calculate the angle of the most suitable straight line among all straight lines passing between the recognized eyes based on the image as the optimal cutting angle.

[0066] In one embodiment, the cutting angle calculation unit (350) calculates an optimal cutting angle using an optimal cutting angle calculation algorithm utilizing the eyes of the seed potato recognized by Mask R-CNN. Specifically, the cutting angle calculation unit (350) generates a first straight line connecting the center of the seed potato image and the center of the recognized eyes. If there are multiple recognized eyes, multiple first straight lines may be generated. For example, if there are three recognized eyes, the cutting angle calculation unit (350) generates a total of three first straight lines connecting the center of the seed potato image and the center of each of the three recognized eyes. The cutting angle calculation unit (350) generates a second straight line as an intermediate angle between the first straight lines, selects up to two of the second straight lines that do not pass through the recognized eyes—that is, do not destroy the eyes—among the generated second straight lines, and determines at least one of the maximum two selected second straight lines as the optimal cutting straight line according to preset conditions. The cutting angle calculation unit (350) transmits the determined optimal cutting line angle value to the cutting operation unit (370).

[0067] In one embodiment, the cutting angle calculation unit (350) may finally select the optimal cutting line based on the number of cases where eyes exist in the cut pieces when cutting using the selected maximum of two second lines. The cutting angle calculation unit (350) may calculate the cutting angle by selecting all of the selected maximum of two second lines when at least one eye exists in each of the cut pieces. The cutting angle calculation unit (350) may select only one of the selected maximum of two second lines when no eye exists in any of the cut pieces, and may calculate the cutting angle by finally selecting the selected one second line when eyes exist in all the cut pieces when cutting using the selected one second line. The cutting angle calculation unit (350) may output a cutting impossible notification message when no eye exists even when cutting using the selected one second line.

[0068] The cutting operating unit (370) can perform a cutting operation of the seed potato according to the calculated cutting angle. In one embodiment, the cutting operating unit (370) can cut the seed potato being transported. To this end, the cutting operating unit (370) may be implemented by including a transport module for transporting the seed potato and a cutting module for cutting the seed potato. The transport module can transport the seed potato to the position of the cutting module. Here, the cutting module may include a blade and a driving means for driving control to rotate and vertically translate the blade. The cutting module can cut the transported seed potato by rotating the blade at a cutting angle and controlling the vertical translation of the blade. The cutting operating unit (370) may include a disinfection module that performs a disinfection function by spraying a disinfectant onto the cut surface of the seed potato or the blade.

[0069] The cutting operating unit (370) can cut the seed potato by repeating the cutting operation according to the number of cutting angles. For example, if there is only one cutting angle, the cutting operating unit (370) can cut the seed potato into 2 pieces by cutting only once according to the cutting angle, and if there are two cutting angles, it can cut the seed potato into 4 pieces by cutting twice according to the cutting angles.

[0070] In one embodiment, the cutting operating unit (370) can cut from the apical portion with many eyes toward the basal portion according to the seed potato cutting guidelines of the Rural Development Administration, and can cut so that 1 / 5 of the basal portion is left intact. Additionally, the cutting operating unit (370) can be implemented in a structure that allows for blade disinfection.

[0071] The control unit (390) controls the overall operation of the seed potato automatic cutting device (130) and can manage the control flow or data flow between the image generation unit (310), the seed potato eye recognition unit (330), the cutting angle calculation unit (350), and the cutting operation unit (370).

[0073] FIG. 4 is a flowchart illustrating an example of an artificial intelligence-based automatic seed potato cutting process according to the present invention.

[0074] Referring to FIG. 4, the automatic seed potato cutting device (130) can generate a seed potato image by photographing the seed potato through the camera sensor (110) in the image generation unit (310) (step S410). The automatic seed potato cutting device (130) can recognize the eyes of the seed potato from the seed potato image through the seed potato eye recognition unit (330) (step S430). The seed potato eye recognition unit (330) can accurately recognize the eyes of the seed potato, which vary in shape, size, and position, by using a Mask R-CNN-based recognition model in the seed potato image.

[0075] Additionally, the seed potato automatic cutting device (130) can calculate the optimal cutting angle among the angles of the cutting line that avoid the eyes of the seed potato recognized through the cutting angle calculation unit (350) (step S450). The cutting angle calculation unit (350) can calculate the most suitable cutting angle of the straight line by considering the number of cases where eyes exist in the cut piece among the cutting straight lines passing between the eyes recognized based on the center of the seed potato image through an optimal cutting angle calculation algorithm.

[0076] In addition, the automatic seed potato cutting device (130) can perform automatic cutting of the seed potato by controlling the operation of the blade with the cutting angle calculated through the cutting operating part (370) (step S470).

[0078] Figures 5a and 5b are diagrams illustrating a seed potato eye recognition model using a mask R-CNN according to the present invention.

[0079] Referring to FIGS. 5a and 5b, the seed potato automatic cutting device (130) is designed to accurately recognize eyes of different shapes, sizes, and positions on a pixel-by-pixel basis using Mask R-CNN. Here, FIG. 5a is the result of training a custom dataset and applying it to seed potatoes, allowing the eyes of the seed potatoes to be recognized by applying a mask.

[0080] The seed potato automatic cutting device (130) uses a camera sensor (110) to photograph the seed potato and performs labeling on the captured image in order to build a seed potato eye recognition model. The labeling work can be performed by using an open-source based mask sense to label the eyes of the seed potato as polygons, as shown in FIG. 5b. The labeled data can be trained on a mask R-CNN to pre-build a seed potato eye recognition model. The accuracy of the pre-built mask R-CNN-based recognition model was verified through the following experiment.

[0081] Experimental Example

[0082] Seed potatoes were photographed from 20 cm above using a Logitech Stream Cam, and after labeling the captured images, a total of 1,118 images were used to train a Mask R-CNN. The labeled data was used to train a custom dataset by loading the Mask R-CNN model into Colab, a cloud-based AI learning environment provided by Google, and the training took a total of 54 minutes.

[0083] The results of running the weight file while increasing the object recognition threshold of the mask R-CNN by 0.05 after training are shown in Table 1 below.

[0084] [Table 1]

[0085]

[0086] Looking at Table 1 above, the F1-Score, which is the harmonic mean of recall and precision, varies depending on the reference value. When the object recognition reference value is set to 0.75, the F1-Score reaches its highest value of 0.926, indicating decent performance. This means that the Mask R-CNN model recognizes the eyes of the seed potato accurately.

[0088] FIG. 6 is a flowchart explaining the optimal cutting angle calculation algorithm according to the present invention, and FIG. 7 is a diagram showing the cutting line calculated through the optimal cutting angle calculation algorithm of FIG. 6.

[0089] Referring to FIG. 6, the automatic seed potato cutting device (130) can photograph the seed potato using a camera sensor (110) and recognize the eyes of the seed potato using a mask R-CNN-based recognition model on the photographed seed potato image. If the automatic seed potato cutting device (130) cannot recognize the eyes of the seed potato using the mask R-CNN, it can output a message indicating that there are no eyes, thereby notifying that the eyes cannot be recognized, so that the seed potato can be re-photographed or the seed potato can be replaced.

[0090] When the seed potato automatic cutting device (130) recognizes the eyes of the seed potato, it can calculate the optimal cutting angle by executing an optimal cutting angle calculation algorithm. The seed potato automatic cutting device (130) generates a first color line connecting the center of the image and the eyes of the seed potato. Here, the first color line may correspond to a red line connecting the center of the seed potato image and the center of the recognized eyes, as shown in FIG. 7 (a) and (b). Figure 7 (a) shows the case where three eyes of the seed potato are recognized, and Figure 7 (b) shows the case where four eyes of the seed potato are recognized, and three and four red lines are generated, respectively. Then, the seed potato automatic cutting device (130) generates a second color line passing through the middle of the adjacent first color line. Here, the second color line may be displayed in blue to distinguish it from the first color line.

[0091] The seed potato automatic cutting device (130) can change a line among the second color lines that does not pass through the eye to a third color line. Here, the third color line is a line that does not destroy the eye when cutting, and may correspond to a purple line that does not pass through the recognized eye as shown in FIG. 7 (a) and (b). The seed potato automatic cutting device (130) can change two lines among the third color lines to fourth color lines so that when the seed potato is divided into four pieces, there is an eye in all four pieces, and calculate the optimal cutting line and angle values ​​by considering the number of cases where an eye exists in the piece when cutting with the selected fourth color line. Here, the fourth color line may correspond to a yellow line as shown in FIG. 7 (a) and (b). The seed potato automatic cutting device (130) can calculate the two fourth color lines as the final cutting angle if there is at least one eye in each of the four pieces cut when cutting using two fourth color lines. If there is at least one eyeless piece among the four pieces, the seed potato automatic cutting device (130) can select only one of the two corresponding fourth-colored straight lines, and if there are eyes in both pieces divided by the selected fourth-colored straight line, it can calculate the corresponding fourth-colored straight line as the final cutting angle. If there are no eyes in both pieces divided by the one fourth-colored straight line, the seed potato automatic cutting device (130) can prevent unnecessary cutting operations by outputting a message indicating that a cutting straight line cannot be selected, thereby notifying that cutting is impossible.

[0093] FIG. 8 is a drawing illustrating the mechanical configuration of an automatic seed potato cutting device according to one embodiment of the present invention, and FIG. 9 is a drawing showing the blade portion in FIG. 8.

[0094] Referring to FIG. 8, the automatic seed potato cutting device (130) includes a frame (810) that forms the body. Here, the frame (810) may be manufactured as a multi-layered rectangular frame having a predetermined width and height using a metal material such as aluminum. For example, the frame (810) may be manufactured with an overall width of 560 mm and a height of 653 mm. The bottom layer of the frame (810) has an acrylic plate installed to form a bottom surface (815). A transfer module (820) for transporting seed potatoes is installed on the bottom surface (815). The transfer module (820) may include a linear rail and a stepper motor that are arranged along the length direction of the bottom surface (815) and transport seed potatoes in the length direction, that is, in a horizontal linear direction, on the bottom surface (815).

[0095] A camera (830) is fixedly supported on one side of the intermediate layer of the frame (810) by a camera holder (835). Here, the camera (830) is configured to correspond to the camera sensor (110) of FIG. 1 and can be mounted on the upper part of the starting position of the transfer of a seed potato placed on the linear rail of the transfer module (820). In one embodiment, the camera (830) may use a webcam, but is not necessarily limited thereto. The camera (830) can capture the seed potato before the seed potato is transferred to generate a seed potato image. The transfer module (820) can be linked with the camera (830) to start the transfer of the seed potato when the seed potato is captured by the camera (830).

[0096] A cutting module (840) is vertically mounted on the other side of the middle layer of the frame (810) from the upper layer toward the bottom layer. Here, the cutting module (840) may be mounted above the end of the transport position of the seed potato placed on the linear rail of the transport module (820). In one embodiment, the cutting module (840) may include a blade (841) located at the lower end, a first driving means (843) that rotates the blade (841) according to the cutting angle above the blade (841), and a second driving means (845) that vertically drives the blade (841). The blade (841) can be periodically disinfected and replaced by creating a gap, as shown in FIG. 9 (a). Additionally, as shown in FIG. 9 (a) and (b), a rubber band can be installed on the blade (841) to prevent the potato from sticking to the blade (841) due to starch after cutting. The first driving means (843) may include a servo motor that controls the rotation angle of the blade (841). The second driving means (845) may include a linear actuator that controls vertical parallel movement to cause the blade (841) to enter vertically toward the seed potato.

[0097] In one embodiment, the seed potato automatic cutting device (130) can cut the seed potato by controlling the transfer module (830) and the cutting module (840) with an Arduino. Here, the Arduino code is written to appropriately parse angle values ​​sent in the form of a list from Python, and a function is input to distinguish between when there is only one cutting angle and when there are two cutting angles, thereby automatically controlling the blade according to the number of cutting angles. As a result of the control, if there is only one cutting angle, the potato can be cut into two pieces by cutting only once, and if there are two cutting angles, the potato can be cut into four pieces by cutting twice. At this time, the seed potato automatic cutting device (130) can cut the potato leaving 1 / 5 of the base, according to the cutting guidelines of the Rural Development Administration.

[0099] Figure 10 is a diagram showing the operation process of the automatic seed potato cutting device shown in Figure 8.

[0100] Referring to FIG. 10, the automatic seed potato cutting device (130) places the seed potato on the placement plate of the transport module (820), photographs the seed potato using the camera (820) (①), and recognizes the eyes of the seed potato using a Mask R-CNN-based recognition model in the captured seed potato image (②). Then, the automatic seed potato cutting device (130) calculates the cutting angle using the recognized eyes of the seed potato (③), and then transports the seed potato to the position of the cutting module (840) via the transport module (820) (④). When the seed potato is transported under the blade (841) of the cutting module (840), the automatic seed potato cutting device (130) operates the cutting module (840) according to the calculated cutting angle to rotate the blade (841) to adjust the angle, and then vertically advances the angle-adjusted blade (841) toward the seed potato to automatically cut the seed potato, leaving 1 / 5 of the base intact (⑤). When there are 2 optimal cutting lines calculated, it is cut into 4 pieces, and each of the cut pieces has an eye (⑥).

[0102] An artificial intelligence-based automatic seed potato cutting device and method according to one embodiment can perform automated seed potato cutting operations by recognizing the eyes of a seed potato to be cut through a seed potato eye recognition model and calculating an optimal cutting line that cuts the seed potato while avoiding the recognized eyes. Accordingly, the mechanization rate at the potato planting stage can be increased to reduce labor and labor costs, and contribute to increasing the profits of potato farmers.

[0104] Although the present invention has been described above with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as described in the following claims. Explanation of the symbols

[0106] 100: AI-based automatic seed potato cutting system 110: Camera sensor 130: Automatic seed potato cutting device 150: Database 210: Processor 230: Memory 250: User I / O Section 270: Network I / O Section 310: Image generation unit 330: Seed potato eye recognition unit 350: Cutting angle calculation unit 370: Cutting operating unit 390: Control unit 810: Frame 815: Bottom surface 820: Camera 825: Camera holder 830: Transfer Module 840: Cutting Module 841: Blade 843: First driving means 845: Second driving means

Claims

Claim 1 An image generation unit that photographs a seed potato to be cut and generates a seed potato image; a seed potato eye recognition unit that recognizes eyes in the seed potato image through a seed potato eye recognition model; and a cutting angle calculation unit that calculates a cutting angle capable of cutting the seed potato while avoiding the eyes of the seed potato, wherein the cutting angle is calculated using a cutting line selected based on the number of cases in which eyes exist in the cut piece among cutting lines passing between eyes recognized with respect to the center of the seed potato image. The cutting operating unit performs a cutting operation of the seed potato according to the cutting angle, wherein the cutting angle calculation unit generates a first straight line connecting the center on the seed potato image and the center of the recognized eye, generates a second straight line as an intermediate angle between the first straight lines, selects up to two of the second straight lines that do not pass through the recognized eye, and determines both of the selected up to two second straight lines as optimal cutting straight lines when at least one eye exists in each of the four pieces cut using the selected up to two second straight lines, and when no eye exists in at least one of the four pieces cut, selects one of the selected up to two second straight lines, and when at least one eye exists in each of the two pieces cut using the selected one second straight line, determines the selected one second straight line as the optimal absolute straight line, calculates the angle of the determined optimal cutting straight line as the cutting angle and transmits it to the cutting operating unit, and indicates that cutting is impossible when no eye exists in at least one of the two pieces cut. AI-based automatic seed potato cutting device. Claim 2 An artificial intelligence-based automatic seed potato cutting device according to claim 1, wherein the image generation unit performs at least one preprocessing operation regarding the seed potato image. Claim 3 An artificial intelligence-based automatic seed potato cutting device according to claim 1, wherein the seed potato eye recognition unit distinguishes and displays the eyes of the seed potato as bounding boxes on the seed potato image using a Mask R-CNN-based recognition model (Segmentation Model). Claim 4 In claim 3, the seed potato eye recognition unit is characterized by pre-constructing the seed potato eye recognition model by using an open-source based mask sensor to label the eyes as polygons on a seed potato image and training a mask R-CNN with the labeled custom dataset. Claim 5 delete Claim 6 delete Claim 7 delete Claim 8 An artificial intelligence-based automatic seed potato cutting device according to claim 1, wherein the cutting operating unit controls the operation of the blade at the cutting angle to perform the cutting operation of the seed potato and re-performs the cutting operation according to the number of cutting angles. Claim 9 In claim 8, the cutting operating unit comprises: a conveying module for conveying the seed potato; and a cutting module comprising a replaceable blade for cutting the seed potato, a first driving means for rotating the blade according to the cutting angle, and a second driving means for vertically driving the blade, characterized in that it is an artificial intelligence-based automatic seed potato cutting device. Claim 10 An image generation step for generating a seed potato image by photographing a seed potato to be cut through a camera; a seed potato eye recognition step for recognizing eyes in the seed potato image through a seed potato eye recognition model; a cutting angle calculation step for calculating a cutting angle that can cut the seed potato while avoiding the eyes of the seed potato, wherein the cutting angle is calculated using a cutting line selected based on the number of cases in which eyes exist on the cut piece among cutting lines passing between eyes recognized with respect to the center of the seed potato image; and a transfer step for transferring the seed potato under a blade. The cutting operation step includes rotating the blade according to the cutting angle to adjust the angle and introducing the blade perpendicularly into the seed potato to perform a cutting operation, wherein the cutting angle calculation step comprises: generating a first straight line connecting the center on the seed potato image and the center of the recognized eye; generating a second straight line as an intermediate angle between the first straight lines; selecting up to two of the second straight lines that do not pass through the recognized eye, and determining all of the selected up to two second straight lines as optimal cutting straight lines when at least one eye exists in each of the four cut pieces when cutting using the selected up to two second straight lines; selecting one of the selected up to two second straight lines when no eye exists in at least one of the four cut pieces, and determining the selected one second straight line as the optimal absolute straight line when at least one eye exists in each of the two cut pieces when cutting using the selected one second straight line; and calculating the angle of the determined optimal cutting straight line as the cutting angle and transmitting it to the cutting operation unit. An artificial intelligence-based automatic seed potato cutting method characterized by including a step of indicating that cutting is impossible if at least one of the two cut pieces does not have an eye. Claim 11 In claim 10, the seed potato eye recognition step is characterized by using a Mask R-CNN-based recognition model (Segmentation Model) to distinguish and display the seed potato eyes as bounding boxes on the seed potato image, thereby providing an artificial intelligence-based automatic seed potato cutting method. Claim 12 delete

Images